Highlights

Kafka: cleanup of unused topics, rebalancing of the cluster to flatten storage and traffic between brokers
WDQS Update pipeline: cleanup updater checkpoints
Dumps: our xml/sql dumps were failing due to an increased memory requirement of 7za in bookworm compared to bullseye. We rolled back the dumps image to one based on bullseye and re-ran all failed jobs
Airflow: Fixed all failing jobs in https://airflow-test-k8s.wikimedia.org. Let's keep the instance green to detect real issues earlier.
OpenSearch:

Migrated translation memory indices on production OpenSearch-on-K8S cluster: now we are fully unblocked for CirrusSearch upgrade, and the Translate extension works faster, because its indices any not getting evicted from disk cache
T425585 Write lightweight OCI-image-based Puppet plans for beta cluster We can now change production OpenSearch without breaking beta cluster.
T426101 Create detailed migration plan for OpenSearch 1.x-2x Migration plan is ready, we will start the migration on Monday

Kubernetes: We hit a hard limit of 32 on the number of workers in dse-k8s-eqiad, so work is now under way to raise this limit to 512, in line with our strategy to migrate more service to Kubernetes.
WDQS v2:

Helped the Wikidata Platform team with getting the new SPARQL backend database to be served by Kubernetes.
T429310 Create a Debian package for the lvmd binary Added repo here. Still hammering out fine details, but should hopefully be done soon.

Presto: Deployed some new configuration options to Presto, with the aid of improving performance and the user exerience for GrowthBook and Superset users.
Data Platform Technical Strategy: Finished drafting A proposed security model for Data Platform services built on Kubernetes and Ceph/S3 and shared this with stakeholders.

OpenSearch upgrade